|
||
150 VIII DIFFERENTIAL CALCULUS
When E has finite dimension n and F has finite dimension m, /'(*<»)
can thus be identified to a matrix with m rows and n columns; this matrix will be determined in Section 8.10. |
||
|
||
Examples
(8.1.2) A constant function is differentiable at every point of A, and its
derivative is the element 0 of jSf(E; F).
(8.1.3) The derivative of a continuous linear mapping u of E into F exists
at every point xeE and Du(x) = u.
For by definition u(x0) + u(x - #0) = «(*)•
(8.1.4) L*f E, F, G be three Banach spaces, (x9y)-+[x'y] a continuous
bilinear mapping of E x F into G. Then that mapping is dlfferentioble at every point (x, y) e E x F and the derivative is the linear mapping |
||
|
||
For we have
[(x + s)-(y+ty\-[x-y]-[x-t]-[s-y]'*ls-t]
and by assumption, there is a constant c> 0 such that \\[s • t]\\ < c * pi • ||f ||
(5.5.1). For any e > 0, the relation sup(||^||, ||f||) « ||(j» Oil < fi/c implies therefore
- (y + 01 -[*•?]-[*•')-[*• y]ll < e life OH
|
||
|
||
which proves our assertion.
That result is easily generalized to a continuous multilinear mapping,
|
||
|
||
(8.1.5) Suppose F = F! x F2 x ••- x Fm is a product of Banach spaces,
andf= (/!,.. .,/J a continuous mapping of an open subset A of E into F. /« orrfer that f be differentiable atx0,a necessary and sufficient condition is that each f, be differentiable at x0, and then f(x0) » (/((jc0)»... f/;(jc0» (wAw JSP(E; F) /j identified with the product of the spaces J5f(E; F,)),
Indeed, any linear mapping u of E into F can be written in a unique
way u = (wl9..., wm), where «ris a linear mapping of E into F|f and we
have by definition \\u(x)\\ = supdlw^z)!!,..., \\um(x)\\), whence it follows
|
||
|
||